Image Classification on MNIST (test) (Accuracy, Loss, and Operational Metrics) benchmark leaderboard · SOTA2 Research

Image Classification on MNIST (test) (Accuracy, Loss, and Operational Metrics)

89.15AccuracyFP32

Evaluation Results

Method	Links
FP322026.07		89.15	0.3491	6.2634	0	0
FP32 + Monotone2026.07		86.95	0.4363	0.3595	99.98	0
QAT Wrap + Monotone2026.07		86.55	0.4426	0.3714	99.99	0.009
QAT Wrap + Monotone2026.07		85.8	0.4694	0.3907	99.97	0.008
PTQ Wrap + Monotone2026.07		84.1	0.542	0.3602	99.95	0.039
PTQ Wrap + Monotone2026.07		84.02	0.5408	0.3587	99.96	0.04
QAT Wrap + Monotone2026.07		82.4	0.5482	0.4585	100	0.012
PTQ Wrap + Monotone2026.07		80.83	0.6709	0.38	99.98	0.041
PTQ Wrap2026.07		75.37	0.7656	3.7467	0	0.793
PTQ Wrap2026.07		74.77	0.7783	3.7592	0	0.796
PTQ Wrap2026.07		64.65	1.008	3.8393	0	0.832
PTQ Wrap + Monotone2026.07		12.5	2.2659	0.1233	99.78	0
QAT Wrap2026.07		11.6	2.4504	5.5541	0	11.719
PTQ Wrap2026.07		11.35	2.2798	0.1417	0	0
QAT Wrap2026.07		10.88	2.4365	5.4734	0	13.922
QAT Wrap2026.07		10.25	2.3037	0.0023	0	0
QAT Wrap + Monotone2026.07		10.25	2.3037	0.0023	0.05	0
QAT Wrap2026.07		9.87	2.5169	5.4968	0	13.638